Method __init__

models_diffusers/unet_3d_blocks.py:1444–1511

(
        self,
        in_channels: int,
        prev_output_channel: int,
        out_channels: int,
        temb_channels: int,
        resolution_idx: Optional[int] = None,
        dropout: float = 0.0,
        num_layers: int = 1,
        resnet_eps: float = 1e-6,
        resnet_time_scale_shift: str = "default",
        resnet_act_fn: str = "swish",
        resnet_groups: int = 32,
        resnet_pre_norm: bool = True,
        output_scale_factor: float = 1.0,
        add_upsample: bool = True,
        temporal_norm_num_groups: int = 32,
        temporal_cross_attention_dim: Optional[int] = None,
        temporal_num_attention_heads: int = 8,
        temporal_max_seq_length: int = 32,
    )

class UpBlockMotion(nn.Module):
    def __init__(
        self,
        in_channels: int,
        prev_output_channel: int,
        out_channels: int,
        temb_channels: int,
        resolution_idx: Optional[int] = None,
        dropout: float = 0.0,
        num_layers: int = 1,
        resnet_eps: float = 1e-6,
        resnet_time_scale_shift: str = "default",
        resnet_act_fn: str = "swish",
        resnet_groups: int = 32,
        resnet_pre_norm: bool = True,
        output_scale_factor: float = 1.0,
        add_upsample: bool = True,
        temporal_norm_num_groups: int = 32,
        temporal_cross_attention_dim: Optional[int] = None,
        temporal_num_attention_heads: int = 8,
        temporal_max_seq_length: int = 32,
    ):
        super().__init__()
        resnets = []
        motion_modules = []

        for i in range(num_layers):
            res_skip_channels = in_channels if (i == num_layers - 1) else out_channels
            resnet_in_channels = prev_output_channel if i == 0 else out_channels

            resnets.append(
                ResnetBlock2D(
                    in_channels=resnet_in_channels + res_skip_channels,
                    out_channels=out_channels,
                    temb_channels=temb_channels,
                    eps=resnet_eps,
                    groups=resnet_groups,
                    dropout=dropout,
                    time_embedding_norm=resnet_time_scale_shift,
                    non_linearity=resnet_act_fn,
                    output_scale_factor=output_scale_factor,
                    pre_norm=resnet_pre_norm,
                )
            )

            motion_modules.append(
                TransformerTemporalModel(
                    num_attention_heads=temporal_num_attention_heads,
                    in_channels=out_channels,
                    norm_num_groups=temporal_norm_num_groups,
                    cross_attention_dim=temporal_cross_attention_dim,
                    attention_bias=False,
                    activation_fn="geglu",
                    positional_embeddings="sinusoidal",
                    num_positional_embeddings=temporal_max_seq_length,
                    attention_head_dim=out_channels // temporal_num_attention_heads,
                )
            )